Impact Analysis using NLP

Introduction

Impact analysis of software change leveraging ML, NLP models from huge number of test cases and requirement documents written in English – Python , Tensorflow / ElMo or equivalent model

Architecture and Flow

Preprocessing

Pre-Processing of Data - Data preprocessing is done to make the text in a form that can be used to analyze and predict the task. A task can be extracting text or keywords from the data file or sentiment analysis or checking sentence similarity. A large number of test cases written in english in a doc file need to be pre-processed before feeding into the model.

USE Model Testing

The USE model provides better similarity index as compared to other models (ELMO, BERT). A simple example showing the cosine similarity using USE model is as follows:

---

Usage of the tool:

1. Download the USE model.

2. Zip file named 5.tar will be downloaded (Please do not change this name)

3. Clone the Github Repository Impact Analysis

4. Create a new folder named model_use in code_files and extract all the files of the 5.tar in model_use.

5. Create two folders named: separate_test_doc & main_document in folder code_files.

6. To create the first embeddings of the test cases -

• Run the python3 automate_cre.py.

• Place the document file containing the test cases (SampleTestCases.docx) in the main_document folder and terminate the process by selecting Ctrl+C, a file named mainfile.data will be created and then run python3 embeddings_update.py.

• After getting the message “saved”, check for the 2 “.npz files” (“embeddings.npz” and “id.npz”) in the folder where all the python files have been saved.

• Search through the document: run the python3 search_engine.py.

• The console will ask for the user to enter either a keyword or a test-id.

• Then, it will ask the user to set up a threshold value. Enter the threshold value in the range of 0-1.

• If the given keyword is not found as per the threshold value, the tool will ask if the user still wants to search some similar matching test-cases (y/n). If yes, then type ‘y’ and enter.

• After the search is completed, you will get the csv file containing the output of the search. The name of the file will be of the format: Similar_to_SearchedString.csv”

• To search for both test-id and keyword, you can run python3 keyword+testid.py and enter the input values asked in the command line.

GUI working

1. Run python3 main_GUI.py (it will take some time in the first time due to loading the model) then you can enter only keyword or only test-id or both. Provide the threshold value as per the need. You can also set the Default threshold value for all the three cases by editing the threshold.csv file.

Output and plot

2. To get a graph plot, run python3 graph_plot.py. Enter the test-id for which you want to get the nearest test cases and press ENTER. The threshold value can be changed by editing the threshold.csv file. A new graph plot will be shown on the screen along with their test-ids.

---

To update/modify the test cases

• If your test cases are in two or more .docx files then there are two steps you can follow:

  1. Delete the existing doc file from the main_document and save it somewhere. Run python3 automate_cre.py and place all the doc files ONE BY ONE (copying one file and pasting it and repeating the same procedure for all the files) in the main_document folder.

  2. Merge all the doc files (along with the existing one) in a single doc file and after deleting the existing doc file from main_document, run python3 automate_cre.py and place the Merged doc file in the main_document folder.

• Now after doing any one of the above steps, terminate the automate_cre.py and run python3 embeddings_update.py. You will get a message of embeddings getting saved.